The article argues for using one big server instead of distributed systems or microservices for most applications, emphasizing that modern servers are extremely powerful and cost-effective. The discussion begins with an overview of a high-end Microsoft Azure server featuring AMD CPUs with 128 cores, 256 threads, up to 8 TB of DDR4-3200 RAM, and high I/O capabilities including 50-100 Gbps network connections. Such a server can handle workloads like 800 Gbps video serving, millions of IOPS on NoSQL databases, hundreds of thousands of web requests per second, rapid Linux kernel compilation, and 4K video rendering at 75 FPS. Cost-wise, renting a similar large server from providers like OVHCloud costs about $1,318/month, while AWS charges a steep $6,055/month, illustrating the cloud premium. Buying an equivalent physical server costs about $40,000, with additional hosting and network charges. While cloud services offer availability and ease of scaling, the “peak load” pricing often makes cloud-native architectures (microservices, serverless) expensive—5 to 30 times costlier than running on a big server. The author suggests that one big server plus a backup in a different datacenter usually suffices for most web services, and scaling vertically is often easier and more efficient than sharding or horizontal scaling. Some objections to one big server are addressed: sysadmins are still needed (as cloud ops), security updates remain necessary, and cloud providers’ high availability comes with complexity and correlated failure risks. Cloud architectures do allow faster development and suit bursty workloads well, but most services do not require such complexity. For caching, bandwidth savings, and latency improvements, CDNs and backups are examples where distribution is essential, and these should be purchased rather than built. Microservices can run on one big server via containers but often add overhead without much gain. In conclusion, the author advocates for a pragmatic approach: use one big server first, keep it simple, avoid unnecessary cloud architecture complexity, and pay attention to costs and operational realities. This approach offers simplicity, lower costs, good performance, and manageable availability when avoiding correlated failures by diverse server types and datacenters.